token-manager
由 kelegele 提供
Coding Agents & IDEs简介
Universal LLM Token Manager - Monitor usage and provide cost-saving recommendations for Kimi, OpenAI, Anthropic.
描述
该技能的功能
Tracks LLM API token usage and costs across providers including Kimi/Moonshot, OpenAI, Anthropic, Gemini, and local Ollama models. Runs scheduled balance checks via cron, records cross-session spending patterns, and outputs cost-saving recommendations based on context usage and model selection.
Direct API balance querying for Kimi/Moonshot and local session recording means you get actual numbers rather than estimates scraped from provider dashboards.
何时使用
- Checking how many tokens a session consumed before closing
- Setting hourly alerts when Moonshot balance drops below a threshold
- Comparing GPT-4o vs GPT-4o-mini costs for a given workload
- Generating a weekly spending report across multiple LLM providers
- Deciding whether to disable reasoning mode on a small task to cut costs
示例工作流
User asks: check my current token usage and get recommendations
- 1
Runs manager.js report with current session token counts, context window usage, and reasoning state
- 2
Queries Moonshot API for live account balance
- 3
Evaluates context usage thresholds and reasoning mode status against recommendation rules
- 4
Outputs cost-saving suggestions such as compacting context or switching to a cheaper model
Session report showing tokens in/out, estimated cost, remaining balance, and actionable recommendations like '/compact' or switching to kimi-k2.5
前置要求
- MOONSHOT_API_KEY environment variable (required for Kimi/Moonshot balance queries)
- OPENAI_API_KEY environment variable (optional, for OpenAI usage tracking)
- ANTHROPIC_API_KEY environment variable (optional, for Anthropic usage tracking)
安装
clawhub install kelegele/token-manager运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。